Format for Qualifier: Each student is to answer 5 questions in total. At least three of those questions must come from the Part 1 and one question must come from Part 2. The remaining question can come from either section.

Quantitative and Qualitative Modeling and Evaluation  

1.

a) In the Proceedings of UIST'99, Hudson and John presented a tool for automatically assigning KLM operators to a GOMS decomposition of a 2-dimensional GUI. Explain what you consider to be the pros and cons of such a design tool if it were extended beyond the initial work of Hudson and John and wrapped into a commercial interactive development environment.

b) More generally, explain why the GOMS family of modeling and evaluation techniques have not been used extensively in the practice of HCI design.

c) Briefly explain another model of human cognition and how it could be used to describe human behavior with an interactive application.

The rich ethnographic descriptions of situated practice and understanding that are the outcome of qualitative research are problematic for systems designers. Discuss the contrast between the practices of ethnographers and the needs of systems designers. To help make your discussion concrete, consider the example of the design of an information appliance specifically targeted for use in a kitchen to support family communications. Your answer should point out the contrast in perspectives for the ethnographer and system designer in understanding effective use of such an information appliance. You should also address concretely in your answer how the ethnographic investigation in this case might be effectively used by system designers.

a) One very strongly supported model of a 2-dimensional graphical user interface is that of a tree of interactor components. Define this model, using examples of GUI research and toolkits that support this model to illustrate your answer.

b) An important problem currently being pursued by many researchers can be referred to as the "scalable interface" problem. Briefly, a scalable interface to an application, such as an interactive card game, is one that adjusts its presentation to the parameters of the device it is presented on. So, for example, a card game such as solitaire would look and feel different on a 17-inch CRT with keyboard and mouse compared to a handheld PDA with only a stylus for interaction. Explain how the interactor tree model of 2-dimensional graphical user interfaces might be exploited to provide such scalability, as well as how this model facilitates higher level tools and application interfaces for providing such scalability. Make sure you differentiate between the facilities provided by the interactor tree and those provided by higher level tools that take advantage of this tree structure. What potential research advances must be made to facilitate this scalable interface problem?

The Model-View-Controller (MVC) Paradigm has been one of the guiding Lights of UI software for the last twenty years. It describes how to build systems that are loosely coupled, can handle multiple interfaces on the same data, and are easily maintained.

a) Why is the global control flow of programs written using MVC-based toolkits sometimes inefficient and hard to understand?

b) Pick TWO non-trivial approaches (ie. something other than "global variables") to dealing with global state and control that have been actually implemented and used in some UI toolkit (can be just a research toolkit as long as it's actually been used). Describe the two approaches, and discuss their pros and cons for use in an MVC-based toolkit. Describe how the two alternatives are different, and how they achieve the same goals. For example, are the alternatives any easier to use and/or more efficient? How do they integrate with the MVC Paradigm?

Designing for small, ubiquitous displays, or SUDs, (e.g., the <10 square inches of the Palm or Helio, touchpads, etc.) is becoming an important activity for HCI designers today, but it seems that some of our HCI Design Methods need to change from the processes used to design desktop applications. In this question, you explore how design activity changes.

a) Pick and describe two applications for small, ubiquitous displays (e.g. using a Palm by doctors to request tests for a patient, using a wall-mounted touchpad to examine the status of the home security system). You will use these applications to highlight similarities and differences with desktop application design methodologies.

b) How will you identify users (and their needs) for your SUDs applications? Are the users different or the same as might be expected for desktop applications?

c) How will you evaluate your SUDs applications? How does the evaluation methodology need to change from desktop application evaluation?

In the Handbook of Usability Testing, Rubin, the author, claims that testing the usability lab under experimental procedures can "expose all major usability deficiencies and their causes for the most common tasks." Do you agree with this claim? Justify your answer with specific examples. What other evaluation techniques would you consider worthwhile? Why?

Consider the following CSCW implementation issues:

1. Making sure that shared data remains consistent.
2. Updating client displays in synchronous collaboration.
3. Alerting users to on-line areas or topics receiving attention by other users.

Now pick one of the following three CSCW application areas:

1. A threaded discussion space
2. An email system
3. An on-line auction site

For each of (a)-(c), describe both how the issues apply to your selected application area and how the issues MIGHT apply. In other words, in this question, we are testing both your knowledge of what currently exists and also your imagination and design capability in extrapolating your knowledge. For example, if you picked (ii) an email system, you should be able to tell us how (a) shared data is typically managed in an email system, but also tell us how shared data MIGHT be handled, e.g., imagine if only a single copy of a message or attachment was actually stored on any server and all recipients actually only stored a pointer to that message or attachment.

For the system you pick, you will be writing SIX small answers: Current and Possible for each of (a) through (c).

Brand-new networking professor Ima Router comes to you with the following problem. She would like to visualize a very large (~100,000 nodes) network in an analysis system that she has built. She wants the visualization to be able to illustrate global topology and relations as well as to show particular attributes of nodes in selected small regions. Discuss a few potential designs that you would tell her about and recommend the best (in your opinion) one for her, arguing its advantages.

Many of the visions and prototypes of ubiquitous computing consist of applications that "do the right thing" based on an understanding of the circumstances of interaction. For example, the Active Badge system caused a phone call for an individual to be automatically forwarded to the phone closest to where the recipient was currently located. This kind of tailorability of an application to currently sensed situations represents a major class of ubiquitous computing application, called context-aware applications.

1. There are two forms of context-aware research activity today. The first tries to attack the software engineering need to ease the burden of building context-aware applications. The second research method aims at building a single, killer context-aware computing application and demonstrate its value to its intended user population in everyday use. Give two examples of each of these research approaches to context-aware computing.
2. In your opinion, which of these two approaches is the better problem for a ubiquitous computing researcher. Justify your answer with specific examples of the strengths and weaknesses of reported research in context-aware computing.